Wheat Growth Diagnosis And Nitrogen Fertilizer Regulation Based On Fixed-wing UAV Imagery

Wheat is one of the most important food crops in China.Rapid and nondestructive acquisition of wheat growth status is of great significance for the accurate management of wheat.In recent years,the fast development of unmanned aerial vehicle（UAV）technology provides a new tool for the rapid acquisition of agricultural information.In this study,wheat filed experiments with different varieties and nitrogen treatments were conducted at Xinghua experimental station for two consecutive growing seasons from 2017 to 2019,including two plot experiments and one field experiment.An advanced fixed-wing UAV platform equipped with RGB and multi-spectral camera was used to obtain wheat field images at the critical growth period,and agronomic indicators such as leaf area index（LAI）,leaf dry mass（LDM）and plant nitrogen accumulation（PNA）were simultaneously obtained.Using the RGB value,spectrum and texture information extracted from the UAV images,LAI and LDM estimation models were constructed based on the fixed-wing UAV platform.Meanwhile,sufficient index（SI）calculated from spectral information of UAV images was used to construct the nitrogen fertilizer recommendation algorithm for wheat jointing stage.This research would provide technical support for crop growth diagnosis and nitrogen management in small and medium-sized planting areas.Firstly,we explored the potential of estimating wheat growth parameters using spectrum and color information assisted by image texture measurements.By calculating spectral color information into vegetation index（VI）and color index（CI）,the quantitative relationships between VI,CI,texture and LAI,LDM were analyzed by simple regression（SR）.The results showed that the correlation between VI,CI and wheat growth parameters is more stable than single texture.While a texture index（TI）,that contains two texture measurements,showed enhanced correlation with wheat growth parameters.Multiple linear regression（MLR）and Random forest（RF）were used to establish estimation models of LAI and LDM using different sets of input variables.Compared with SR and MLR,the RF model combined with VIs,CIs and TIs improved the estimation accuracy of LAI and LDM,with the validated R2 for LAI estimation model increased from 0.66-0.71 to 0.76-0.81 in three different wheat growth stages,and the validated R2 for LDM model validation increased from 0.65-0.71 to 0.75-0.78.Furtherly,we studied the feasibility of using spectral information from UAV images to guide wheat variable rate nitrogen fertilizer topdressing.The relationship between VI and plant nitrogen accumulation（PNA）of wheat was analyzed using data of wheat plot experiment with different varieties and nitrogen application rates,and the optimal VI,RESAVI,was selected to establish the PNA estimation model,with R2 reached 0.78.Then the relationship between total nitrogen application rate and wheat grain yield was analyzed to determine the optimum nitrogen application rate（NOPT）.Sufficient index（SI）was obtained by comparing the spectral value of the area to be fertilized with that of the area under sufficient nitrogen supply.The SI-based nitrogen fertilizer regulation algorithm was constructed using key parameters such as PNA,NOPT and SI.Finally,the nitrogen regulation algorithm was verified on the wheat field experiment.The results showed fixed-wing UAV imagery combined with SI algorithm could effectively guide wheat N topdressing according to wheat growth status,with preferable grain yield performance,nitrogen utilization efficiency and economic benefits.This study demonstrated that texture information can be used as supplementary information to effectively broaden the dimension of UAV image data and to improve the accuracy of crop growth monitoring.In addition,the fixed-wing UAV can effectively diagnose the nitrogen status of wheat and guide the application of nitrogen fertilizer for small to medium size of farmlands.